Backup has always been a key part of storage operations. In the early days, punch cards, printouts, punched paper strips and magnetic tapes running on refrigerator-sized drives duplicated everything that resided on mainframe drives. For the past several years, however, backup has taken a backseat in terms of media attention, overshadowed by stories of mushrooming capacity, SAN, NAS and virtualization.
Despite the lack of hype, however, backup technologies and methodologies have continued to develop. While it is difficult to isolate the backup market from the overall storage management market, Gartner Inc. (Stamford, CT) estimated the market at around $1.5 billion last year, a nine percent increase over the previous year. Veritas claims 41% of the US market followed by Computer Associates (CA) and Legato Systems; CA claims number one spot in Europe with a 45 percent share.
Recently, recovery rates in particular have been receiving a lot of attention. According to Michael Karp, a storage analyst for Enterprise Management Associate (EMA — Boulder, Colorado), "99 percent of IT sites in North America don't know how much of their backup data is actually recoverable."
But recovery efficiency is a moot point if you can't even conduct a thorough backup. The traditional approach of overnight or weekend backups is proving inadequate to the 24/7 ecommerce age. With systems operating around the clock, backup windows have diminished or disappeared altogether in some cases.
Exacerbating this problem is the rapid expansion in the volume of data being stored. Not only is more business being transacted and stored electronically, but the transition from text documents to graphics-rich and multimedia files greatly expands document size. Then there is the complexity caused by using multiple device types and platforms throughout the enterprise.
That is what PCMall in Torrance, California ran into when it supplemented its mix of Windows and Unix servers with some Linux boxes for ecommerce, NAT, DHCP, DNS and IPChains. Although the company was pleased with the way operating system itself performed, it found that there was no adequate backup utility that would backup all its machines. That only recently changed when CA came out a backup product that supports all three platforms.
"Initially, we were simply tar-balling the Linux systems up to tape or we'd use AMANDA or some other open source alternative which would work well for backups but there was no way of unifying our environment," says Christopher Swadish, Director of Operations at PCMall. "Data is the bloodline of our company and having complete control over this factor is extremely important."
But, backing up is only half of the equation. Restoring files is just as problematic. With so many backup tapes to manage, it becomes difficult to find the ones you need. Tapes breaking, not containing the data desired and restores not working are frequently voiced as problems by storage administrators.
"Research shows that companies are unable to restore 20 to 25 percent of their backed up data whether the problem is due to operator error or faulty equipment," said EMA's Karp.
He reports, however, that backup finally is becoming more automated and less intrusive i.e. the backup process doesn't interrupt or degrade other processes and can provide extremely rapid, low impact recoveries. The overall vision being embraced by the large backup vendors is to create software that enables IT to push a few buttons and conduct an enterprise wide backup, one that can be recovered successfully when needed.
Rather than being propelled by one technology, this dream is being fueled by advances on several fronts:
Disk to Disk Backups - As the price of hard drives has plummeted in the past few years, it has reached the point where it can be economical to directly backup to another disk rather than to tape. Disks greatly accelerate the backup process and data can later be transferred to tape for offsite archival storage without impacting ongoing operations.
Snapshots - Rather than performing a complete backup, some types of snapshot capture changes made to files. This reduces the amount of data that needs to be backed up at any point in time and cuts down on the amount of data traveling over the network. Such snapshots don't replace full backups, but they do provide a means to backup data more frequently without producing an excessive resource drain. And they offer an escape from the dilemma of ever-shrinking backup windows.
Fibre Channel over SONET - Backing up data to a secondary datacenter provides more complete protection than backing it up on site. But Fibre Channel, although it provides high data transfer speeds, has a limited range of only about ten kilometers. In the case of a major disaster, this is not enough. Newer methods of converting a Fibre Channel signal to SONET allow the data to be backed up over a distance of thousands of miles, while also cutting bandwidth costs.
These and other breakthroughs are contributing to the goal of automated multi-platform backups that can be recovered on demand. And with the PC no longer the focal point of enterprise information, backup products are being stretched to the limit as they struggle to cope with data coming from PDA's, laptops, PC's and mainframes. As yet no one product does all of this.
However, with the mix of technologies that have emerged in the last few years, the job can be done. And the drive to fulfill this demand, though it may not have returned backup to center stage, has at least enabled the technology to begin to emerge from the long shadow of the architectural side of the storage industry.